Guest: G S Madhusudan – Cofounder & CEO of InCore Semiconductors
Episode Summary:
What will it take for India to become a true global power in the semiconductor industry? In this episode of Chai & Chips, I sat down with GS Madhusudan, a veteran technologist and the co-founder & CEO of InCore Semiconductors, one of India’s pioneering RISC-V processor IP companies. Madhusudan provides a masterclass on the Indian semiconductor landscape, breaking down the complex ecosystem into distinct verticals—from design and product to foundry and equipment. He offers a pragmatic, no-nonsense roadmap for progress, emphasizing the critical need for measurable goals, government-led demand creation, and a fundamental shift in mindset from mere local production to true self-belief (Aatmanirbharta). This is an essential conversation for anyone looking to understand the real challenges and immense opportunities in building India’s deep-tech future.
YouTube episode link:
Key Insights:
In a compelling and pragmatic discussion on Chai & Chips, GS Madhusudan (Madhu for short), Co-founder & CEO of InCore Semiconductors, laid out a clear-eyed blueprint for India’s journey toward semiconductor sovereignty. Moving beyond high-level view, Madhusudan provided a granular, actionable framework for founders, policymakers, and investors. His insights reveal an ecosystem at a critical inflection point, where strategic choices made today will define India’s role in the global technology landscape for decades to come.
Here are the key insights and takeaways from the dialogue.
1. Deconstruct the Ecosystem: A Vertical-by-Vertical Approach
Madhusudan argues that a single, monolithic “semiconductor strategy” is meaningless. The industry is not one entity but a collection of at least five distinct verticals: Design, Product, Packaging, Foundry, and Equipment. Each is at a different stage of maturity in India—from a well-established design ecosystem to a nascent equipment industry.
His prescription is simple yet profound: conduct a capability audit for each vertical and set separate, concrete, and measurable goals. Instead of a vague 2030 vision, the goal should be specific: “Our fabless guys need to turn out 100 million chips by 2030.” This targeted approach allows for focused effort and prevents the complexities of one vertical from stalling progress in another.
2. The “25 by 25” Vision: A Measurable Goal for a Nation
To anchor India’s ambition, Madhusudan proposes a powerful and memorable goal: the 25 by 25 vision. The objective is for India to capture 25% of the global market for semiconductors costing below $25. This targets the high-volume, mature-node segment where India can realistically build a dominant position. This is not just about manufacturing; it’s about owning the design and product value chain. This vision provides a clear, quantifiable benchmark for success that can align the entire ecosystem.
3. The True Meaning of Aatmanirbharta: From Local Production to Self-Belief
One of the most powerful themes was Madhusudan’s redefinition of Aatmanirbharta (self-reliance). He stressed that it goes far beyond the simplistic idea of local manufacturing.
“Aatmanirbharta is not just doing stuff locally, it’s believing in yourself. People miss that aspect of it.”
True self-reliance is about having the courage to conceptualize, the confidence to innovate, and the resilience to fail and try again. This cultural shift from a service-execution mindset to a product-ideation mindset is the fundamental prerequisite for building an indigenous technology ecosystem.
4. Building the “Boring” Foundation: The Unsung Heroes
Success in semiconductors isn’t just about glamorous, cutting-edge technology. Madhusudan emphasizes the critical importance of building the “boring” but essential product ecosystem: supply chains, distribution networks, marketing, field application support, and RMA processes. He uses the analogy of building a base camp before attempting to summit Mount Everest—it’s a long, arduous process that takes years but is non-negotiable for success.
This requires courage, which he notes comes from experience and a deep network. He highlights the crucial role of unsung heroes like EMS (Electronics Manufacturing Services) companies and distributors, who provide the market intelligence and logistical backbone necessary to turn a chip design into a commercially viable product.
5. The Government’s Role: Being a Demand Facilitator
Madhusudan is unequivocal about the government’s role. Its most powerful tool is not equity funding but demand creation. By implementing strong mandates for local procurement in strategic sectors (meters, smart cards, government hardware), the government can guarantee a market for Indian-designed chips. This de-risks the venture for private capital, making VC funding far more accessible.
Government’s strength lies in being a strategic customer and a facilitator, creating a simple, predictable regulatory environment that encourages innovation rather than stifling it with red tape.
6. RISC-V: The Democratic Enabler of Choice and Innovation
RISC-V is a game-changer because it is an open and democratic instruction set architecture (ISA). Unlike proprietary ISAs like Arm or x86, it provides “choice.” This lowers the barrier to entry for startups and allows for deep customization and innovation without prohibitive licensing fees. While the open nature presents challenges like potential fragmentation, it also harnesses the collective brainpower of a global community, accelerating progress in a way closed ecosystems cannot. For India, leveraging RISC-V is a strategic imperative to build a sovereign and innovative chip ecosystem.
7. A Call to Action: Humility, Patience, and Hard Work
The journey is a marathon, not a sprint. Madhusudan calls for a “large dose of humility,” especially from industry veterans, who should see their role as laying the foundational “bricks in a wall” for the next generation. For aspiring founders, his advice is clear: know your market well. Spend time understanding customer needs before perfecting the technology. The ecosystem desperately needs more marketing and business-savvy leaders to complement its deep technical talent. Ultimately, building a successful deep-tech company is 99% hard work—a relentless focus on the details that turn a magical idea into a tangible reality.
Connect with our Guest:
InCore Semiconductors: https://incoresemi.com/
G S Madhusudan on LinkedIn: https://www.linkedin.com/in/gsmadhusudan/
Follow Chai & Chips:
Subscribe on YouTube(Chai & Chips - YouTube), Spotify(Chai & Chips Podcast | Podcast on Spotify), Apple Podcasts(Chai & Chips Podcast - Podcast - Apple Podcasts) or your favourite podcast platform to never miss an episode.
Leave a review—it helps us reach more listeners!
Follow Prakash on LinkedIn(Prakash Mallya | LinkedIn) and X(Prakash Mallya (@PrakashMallya) / X) for updates.
Episode Transcript
Speakers:
Prakash Mallya (Host)
G S Madhusudan (Guest)
(00:04) Madhu: Semiconductor has like about five to six verticals. Correct? You have the design ecosystem, which exists courtesy MNC companies, courtesy local companies, we are very mature. Then you have the product ecosystem, which is almost non-existent. But large companies like L&T Semi, IVP, Axero, and smaller companies like Mindgrove are coming about, NetraSemi that is there, right? So that’s the start of a revolution. Then you have the packaging industry which is starting to come, investments are getting made. Then you got the foundry industry, which is just taking baby steps. Then you have the equipment industry, which is almost non-existent. I mean who supplies the equipment to the foundry, right? So, each of those verticals, my standard advice is, look, do a capability audit and set reasonable goals for that particular sector.
(00:56) Madhu: So, saying we will reach some place in 2030 is meaningless. You should say, okay, our fabless guys need to turn out 100 million chips by 2030. That’s a goal for the fabless sector. But always wanted to do processors, always wanted to do vector processors for some reason. And so InCore is almost a bucket list item. But I never trained as a processor guy.
(01:19) Madhu: Aatmanirbharta is not just doing stuff locally, it’s believing in yourself. People miss that aspect of it. India probably needs to have 25% of the world’s market in semiconductors costing below $25. For me it’s a 25 by 25. One fourth of the market should be chips designed in India.
(01:44) Prakash: Welcome everyone. Welcome to Chai & Chips. This is Prakash and today we have with us Madhu. Madhu for short is the Co-founder and CEO of InCore Semiconductors, one of the first processor IP companies of India. He was also the Co-founder and still is of Shakti, our RISC-V, India’s RISC-V venture, which has done tremendously well and we’ll talk about it. He has held a variety of different roles across technology and startup businesses and R&D organizations around the world. He’s an advisor to Indian government and most recently he has been the advisor to Government of India in AI task force as well.
(02:29) Prakash: I wanted to quote a quote from Madhu in some past conversations just to give you a measure of the person himself. So he said, “We had Green Revolution to make sure we grew our own food. We need Silicon Revolution so that we can feed our computing needs ourselves.” So with that context, ladies and gentlemen, Madhu, welcome to Chai & Chips.
(02:56) Madhu: Thanks a lot for inviting me. I’m really glad to be here.
(03:01) Prakash: So, I’ve seen and read and listened to many, several conversations you have done in the recent past and you have been very clear and sometimes unapologetic about the semiconductor strategy India should pursue. So, I wanted to dive deeper into this conversation about the why, what, and how of the strategy of building the indigenous and strategic chip ecosystem for India. I hope that is okay with you, Madhu.
(03:32) Madhu: Oh, that’s perfectly fine.
(03:34) Madhu: So, the problem is everybody says strategy, right? As if it’s a single problem to be solved. Semiconductor is like about five to six verticals, correct? You have the design ecosystem which exists courtesy MNC companies, courtesy local companies. We are very mature. Then you have the product ecosystem, which is almost non-existent. But large companies like L&T Semi, IVP, Axero, and smaller companies like Mindgrove are coming about. NetraSemi that is there, right? So that’s the start of a revolution.
(04:10) Madhu: Then you have the packaging industry which is starting to come, investments are getting made. Then you got the foundry industry which is just taking baby steps. Then you have the equipment industry which is almost non-existent. I mean who supplies the equipment to the foundry, right? So each of those verticals my standard advice is, look, do a capability audit and set reasonable goals for that particular sector.
(04:34) Madhu: So, saying we will reach some place in 2030 is meaningless. You should say, okay, our fabless guys need to turn out 100 million chips by 2030. That’s a goal for the fabless sector. The design industry should be designing along with similar to Broadcom and Marvell of saying at least if we don’t design the top of the line uh AI/ML training or ML chips, we should be at an 80% level and we should have at least two of the majors say AWS and Google as our customers.
(05:06) Madhu: In the equipment industry you can say that look, we need all the base gases and all of that and we should at least be supplying equipment at a 90 nanometer level. In the fabs you should say okay, we need, do you need silicon carbide fabs? Yeah, we need because we are, we consume a lot of power parts and all that, right? But we also need mixed signal fabs and up to 40 nanometer is okay. I’m satisfied with that, right?
(05:30) Madhu: So each vertical in the semiconductor needs a certain vision, it needs a roadmap, we need to figure out how much of public private contribution needs to be there. So, I would prefer to see a task force and mission mode for each vertical and let each one of them take their own course. Talking about semiconductor as a whole does not do justice to what is a very vast industry.
(05:52) Madhu: My prescriptive advice, I stick only to fabless and IP because that’s all I know, right? I’m not going to tell a packaging guy to do this and stuff. I know they do wire bonding and stuff and that’s all I, I get it, right? Whether we should do 2.5D, we should do 3D, it all depends on what devices you make. And same thing with the fabs also, right?
(06:12) Madhu: So we need to split the problem into six different problems, have a roadmap, and have concrete vision. The key thing is to have measurable goals otherwise you can’t measure progress. The fundamental problem we have is what are we supposed to do say in the fabless by 2028? Can we take five chips? Smart cards, electricity meters, IP camera, handheld terminal chips, a general-purpose controller chips and say that we will be shipping 2 million of these by 2028, each of these. 10 million total, right? Doesn’t matter. It can be a small goal, it can be a large goal. But we need concrete goals, we need measurable goals and we need to set such goal for each sector.
(06:54) Madhu: I don’t see such a comprehensive approach yet, but everybody’s strapped for bandwidth. I’m not going to blame MeitY on this, okay? MeitY is strapped for cash, MeitY is strapped for bandwidth and so there’s only so much you can expect the government to do. I frankly think it is up to the private sector to step up.
(07:12) Prakash: So let me try and unpack all of what you said during the course of this conversation, right? So the first part is something you have mentioned in the past and you alluded to it in your comments as well, is we are as a country, despite having great amount of chip talent, 20-25% of the chip designers based in India of the global supply, we are on the execution end of the supply chain. So you have an IP or an idea coming from abroad, you execute on it. What is the gap? What makes us stop short from defining our own ideas versus just executing on the ideas that come from abroad?
(07:56) Madhu: In one simple word, courage. The courage to conceptualize a product and push through does not come from raw courage. It also comes from experience. So when all the MNCs came here, we basically designed products based on somebody else’s specs. We might have suggested an architecture change and all of that. In the architecture itself we have progressed. So, you do have designers who can innovate an architecture, that’s not a problem. Intel, AMD, TI, Analog Devices, to name a few, have created very forward-thinking designers. But at a product level, we never had product visionaries, quite simply those roles were not needed in India.
(08:36) Madhu: So even the best of sales guys were always given a line card from the US and said, say go sell this. Now I know all, most of them I know, okay, I’m pretty sure they’re all listening. They can ideate. They were never given a chance to ideate. So, if the local head of sales for say one particular company says, “Hey, this is a great product for India,” I don’t think his idea was entertained. Right? So, we never gave a chance for people in the sales and marketing organizations in India to create products, to fail in creating products. Doesn’t matter, right? You give it four tries, you get four failures, the fifth one will succeed, right? That’s what I said, it comes with experience. Confidence doesn’t come because you’re arrogant. Confidence comes because you’ve built that network.
(09:27) Madhu: Look, the product ecosystem is very network-based, okay? So if I want to do a product, the first thing is I’ll go to a disti. I’ll ask him, “Look, I want to do this chip.” He’ll say, “Madhu, I sell 5 lakh a year. I sell them for about $2.” Okay? Now the first thing he’ll say, ignore the auto guys, they need qual and all that, they will never touch a startup. But I do have guys who are doing very non-critical, like this door management system and stuff who’d be willing to take a gamble on you. I think I can sell about 100,000 parts at about a $1.80 because you need to be about 10-15 cheaper, right? This conversation that happens, that is what gives me confidence in launching a part.
(10:08) Madhu: So we need an ecosystem of product creators, product marketing guys, and a distributors who are willing to talk to you. This network takes time to build. It doesn’t come in a day. Okay? Now the distributor also won’t take a risk on you unless you’ve shipped a few parts, otherwise he will burn his fingers, correct? So he needs to know that you got a product ecosystem that not only can deliver good parts, you got proper RMA, you got proper field application support, you got the software stack, uh you got the supply chain done, you have reasonable MOQs, you have a tie up with a foundry that if suddenly an order picks up you can deliver. All of this matters, correct? This ecosystem of creating a supply chain is almost non-existent in India.
(10:52) Madhu: Okay. So, if you suddenly say we need to become a product nation, you can’t create all of this, right? My standard example is you don’t attack Mount Everest, right? You build the base camp. And it takes about 10 years to build the base camp. Building a base camp is boring stuff. Most of what is involved in this is extremely boring stuff. It is not glamour technology, right? It’s run of the mill stuff. It is no different from setting a supply chain to supply rice. Very few people want to do this. But there is a reason why Tim Cook is the CEO of Apple, right? The dude has got supply chains figured out. It happens in the armed forces also. I think Patton or somebody said, professionals talk about logistics, amateurs talk about strategy. Correct? It’s logistics, it’s supply chain, it’s boring, mundane stuff that we need to invest in. There’s no shortcut to that. And that’s what we need to do.
(11:44) Prakash: Let me ask you, for the benefit of your audience, MOQ is minimum order quantity. And a follow-on I had was, I agree with what you said, and you started your career in HCL if I’m not wrong. I started my career in Wipro. Both of them were hardware companies at one point in time. And we lost a lot of IP or product capability over time. What do you think went wrong?
(12:13) Madhu: Oh I know exactly what went wrong because I’ve had those conversations umpteen times with all the senior guys both at Wipro and HCL. Easy money. Easy money. As a team, I’ll tell you for example what happened at HCL, right? HCL did everything. Correct? Hardware local, operating system local. The first HCL operating system Alpha was written from scratch in Chennai. And I remember when I joined, the operating system was being deprecated. The last copy of the operating system was being sold to the rabriwala (the paper guy). I was almost in tears. This is an operating system that you wrote from scratch, you’re not maintained copies of it, and the last printed copy is being sold as for paper recycling. That’s the value. I keep asking HCL, do you guys have a museum where all the stuff that HCL has created is there? I wish they were. Same thing for Wipro, same thing for ECIL, like five companies that did great stuff, correct? So in HCL what happened? So the reason I went to the US was actually to port Sybase onto HCL hardware. So the initial HCL teams that went to the US was actually on a collaborative mission. Then Sybase found out that we had great engineers and said, “Hey, why can’t they help us on other things and all of that, right?” So slowly they smelt blood. But HCL did the ethical thing. HCL didn’t say HCL. We were all placed in a company called Hindustan Manpower Training Services. So HCL said, if I’m going to do services, I won’t be a hypocrite. I will call it services. HCL R&D was kept separate. The body shopping, which is what it is, was kept separate, right? Then they did the, but HCL had a problem. Look, the ecosystem in India was not good enough to create a world class hardware.
(14:03) Madhu: So when HCL tried to ramp up, there were quality ramping up problems and all of that. And so Arjun Malhotra went to the US to start selling HCL boxes, not to create services. HCL America was created as an offshoot for selling hardware in the US. It’s not as if HCL didn’t try. So please don’t blame people, right? It was difficult. Then the HCL-HP merger happened and they kind of like broke apart and stuff. And it became tough to compete with the foreign hardware companies and the government I don’t think was very supportive at that point of time. So, it was a combination of easy money from services and a desire to survive that made all these companies go into services.
(14:44) Madhu: Hey, by the way, HCL owns Lotus Notes now. HCL owns Ingres. Okay, so they they are a bonafide product company. And same thing with, I mean TCS sells TCS BaNCS, Infosys sells Finacle. Why do you keep calling them services companies? See, you need to make a distinction between services as a business model and services as a talent model. If a UK company builds you a Formula One car, you don’t call that company a services company. You call them a tech company that can build you a Formula One car. If somebody builds you a rocket custom built, you call them a great tech company. So why do you persist in calling HCL and Wipro and Infosys and TCS service companies? I think you’re doing injustice to the talent that is there in those companies. They have chosen services as a business model doesn’t mean they’re not investing in technology.
(15:45) Prakash: That’s fair. So let me ask you a question on the comment you made earlier about having the courage and conviction of the market and diving in to create a product. So, wind the story back now versus how it used to be on hardware, when we had a set of products but we somehow lost the capabilities to build products here on the hardware side including semiconductor. And today, what are the systemic issues that you see or are there things that we could do better for aspiring founders to invest more of their energy in building products, especially chips and hardware products out of India?
(16:30) Madhu: Look, I advise a lot of startups. I work with a lot of them. I don’t spend much time, occasional conversations, right? Every startup is a techie-led startup. I can barely maybe one or two startups which come from a marketing background. InCore is one of the few companies where one of the co-founders has got significant management and marketing experience and IP creation experience, right? But experience like ours is pretty rare. So, most of them are very deep tech startups, excellent technical talent, but most of them have no clue about the market. Can’t blame them, that’s not where they came from. Correct?
(17:12) Madhu: So I don’t know what the answer is. So, the biggest problem we have is it’s very, next to impossible to hire a good VP of Marketing. We were very fortunate to get Deepak Sahu, our Head of Marketing. IIM-B guy, spent 10 years at Hyperverge and other companies and stuff, right, has done marketing around the world for product companies, right? He may not be a semiconductor guy, but he knows how to go to market and create and all of that, right? So he’s a good find. But people like him are tough to find. Right? And that is a lacuna. So my standard sound bite answer is, the IITs have done the job, it’s now time for the IIMs to do their job. So, we even toyed with that. In fact, I talked to folks at IIM Bangalore. We probably need to create workshops to help create VPs of Marketing. Do case studies on how to launch a product, how to identify market opportunities, how to set up supply chain. How do you break into a market where there are 20 MNCs who have cornered the market? Where do you get the courage? How do you find a spot? You just need to find that one weak spot, right? And then keep driving that nail through it. That is where the courage comes.
(18:29) Prakash: Does the Valley train the founders any differently? Because they seem to be doing it so much better for several decades and great companies have come out of them.
(18:38) Madhu: No, the Valley is a comprehensive ecosystem. The Indian ecosystem is very lopsided. This is a very techie ecosystem. And it’s a cultural problem also, right? The Valley values risk-taking. Indian culture doesn’t. There, failure is a badge of honor. Here it’s a mark of shame. So, the culture conspires against you. It’s changing, it’s changing. Right? Look, I’m not blaming anybody. This is not a blame game. These kind of cultural shifts take time and you and I can keep pontificating till the cows come home. Things will happen in their sweet time.
(19:17) Prakash: And I feel at least my parents would prefer for me to take a salaried job. But I think the next generation, especially the number of founders coming out of India in any sector is so much higher, order of magnitudes higher than when I came out of college. So definitely things are improving.
(19:38) Madhu: No, my dad was just the opposite, right? I still remember an episode. He suddenly came and said, I don’t know if you’ve heard of Travancore Titanium. There are a PSU and dad was an advisor to the government and all that. So he used to handle a lot of BIFR cases and all that sick industries and all that. So he said, “Travancore Titanium is gone bankrupt and stuff. Do you kind of want to do a leveraged buyout and take it over?” I said, “Why on earth?” He said, “You kept talking about titanium.” Titanium is material science, is of interest to me purely as a hobby, academic interest. So he just assumed that if a titanium company came up for sale, I would dump everything and run that. So, he didn’t think twice. He said, “Look, should I talk to some of my friends and we can see if we can take it over?” I said, “Dad, that just was a hobby thing, I’m not that serious, right?” So, he never discouraged me. If I said that look I’m going to start a company, he will simply say yes and saying that if you need help getting a good board of directors, let me know, I’ll get you. If you need, right? He would assume that if I thought of something that I would have thought it through, I would be treated as an adult and his only thing was go do it properly. That’s all. Nothing else. Never, not even a sentence of, he would never even say have you thought it through. He just assumed that whatever if I say, I’m serious about it, right? That kind of home atmosphere helped a lot.
(21:04) Prakash: 100% agree. Yeah. And from what I’ve read, he also advised you or coached you on how to run a business based on his experience of running a business.
(21:15) Madhu: Oh, it’s far more than coach, right? I’m a third-generation CEO. So, he taught me PE investing when I was 14, read balance sheets, how to do projections, how to evaluate a company, how to evaluate a company in half an hour, look at the numbers and stuff and all that, right? And then he would also tell me how his board meetings went, how to handle board conflicts, how to hire the right HR, and also more importantly, how to delegate. And these were explicit instructions. In fact, the only thing he would ever gift me year after year was the latest ROC notifications, tax law changes, and GST changes. I don’t think he gave me ever anything in my life other than that. I mean, I’m not seen him, I don’t remember him giving me a shirt or anything like that, right? Promptly every year anything related to corporate affairs, the necessary documentation, he would send it across and say read. That’s all.
(22:08) Prakash: Well, in the end, that added huge value to you in terms of everything you’ve achieved.
(22:14) Madhu: That is for other people to judge whether I’ve been successful as a CEO is in various vendors and stuff. I will not venture an opinion, right?
(22:22) Prakash: Achieved things that not too many people do in terms of processor IP company, how many exist in this world? So, fantastic job.
(22:31) Madhu: That was an interest for me right from college. So, I was starting to dabble with 8-bit processors all of that. I’m an electronics major by training. I’m not a software guy. But I came to HCL because there’s no other job available. TCS wanted to offer me some COBOL product job which I was not interested and living in Bombay on 3,000 rupees a month back in that time was not keen….I didn’t want to live in Bombay. Chennai was, we had our home and all that, so that was the case. So I took that job. But always wanted to do processors, always wanted to do vector processors for some reason. And so InCore is almost a bucket list item. But I never trained as a processor guy. I am eternally grateful to Neil Gala, Gautam Doshi and Arjun Menon my founders in Shakti. Look, I can’t design processor on my own, right? Together we created something that let me scratch my itch also. I’m grateful to them.
(23:32) Prakash: So, to go back to what we are talking about, product conceptualization and IP. And you have at length talked about how in different segments and drawn fantastic parallels from those segments like two-wheelers, garments, and even whiskies, where India has put the country on the map by developing great IP, conceptualizing it locally and scaling it globally. So, what things could you draw from it or one or two areas that can apply to semiconductors, chips, manufacturing as well that we could take forward as Made in India products?
(24:12) Madhu: Have you heard of Windmill fans?
(24:15) Prakash: Yes, I have.
(24:17) Madhu: So, I got like about four of them, right? You know where their background is, right? They used to do the railway fans that you see on the top of, right? That that old fans that you see, that cage fan in the old train. So that is a family that’s been doing it almost since independence. So, this is the second or third generation which said, “Why don’t we do designer fans?” And their fans start at 20,000 rupees minimum. And go up to a lakh plus and all of that, right? Designer fans, whatever material you want, you can give a sketch and they’ll do a fan for you on that, right? So, they ideated, they felt there’s a need in the market. It’s a very high-end market, they export a lot. Correct? So even in chips, you simply need to sit down, just go to your nearest consumer electronic store and look at all the products that Croma is selling. Take a guess as to what chips are there inside. Correct? Then go talk to the distributor to see what is the price, what are the volumes and stuff, right? You need to educate yourself. You need to have a feel for the market.
(25:15) Madhu: Again, some of this I learned from my dad, right? I mean running the company he would always delegate, but he would spend a lot of time in the Sataa Bazar in Bombay, actually used to spend a lot of time with, I think Dhirubhai Ambani’s brother, the guy who was in the cotton textile trade side of it, and stuff. So what do you do there? He says, “No, Madhu, I just feel the pulse of the market because I’m buying cotton from the mill, so I need to have a pulse of the market to figure out what price to buy, when.” So being a public sector CEO, he would actually buy on the spot market without a contract or without prior approval. The government used to scream at him and he said, “Look, you can’t negotiate. Cotton market you’re there in the market, suddenly you see a price drop, you sign a contract and you buy immediately because a written contract won’t get honored after a period of time.” Correct? So you need to have a feel for the market. See, we take a too a techie an approach to products. Products are products. Doesn’t matter if you’re selling t-shirts, you’re selling two-wheelers. Fundamentally they’re all the same, right? You need to have a feel for the industry. And a lot of us are coming into the industry knowing only one part of the industry and that’s why we are failing.
(26:25) Prakash: Yeah. I’m an engineer, but my background is sales, and what you said, absolutely resonates with me. The market savvy and pulse of the market is so important to build products that customers want, not what you think customers will buy.
(26:41) Madhu: So what I do is I often go to Ritchie Street which is the equivalent of Lamington Road, right? I have a friend. And I just sit by his side, he’ll be in the counter, and see who buys. Whether Gigabyte sells more, ASUS sells more, do people ask for more PCIe slots, less PCIe slots, how much GB of RAM do they buy? Correct? Which size monitor? Do they buy backlit, are people going for OLED and stuff? You just sit for three hours and just watch people ask. You’ll get a feel for what happens in the PC market.
(27:10) Prakash: So that’s a good segue into taking somewhat of a short break and doing a set of lightning round questions. Are you okay with it?
(27:20) Madhu: Yeah.
(27:21) Prakash: Should be fun. Okay, so let me ask you one word to describe the current state of India’s semiconductor strategy.
(27:32) Madhu: Early exploration.
(27:34) Prakash: Those are two words.
(27:36) Madhu: Exploration.
(27:38) Prakash: What’s the most overhyped aspect of the global semiconductor industry right now?
(27:43) Madhu: AI chips.
(27:44) Prakash: If you could instantly solve one non-technical problem in India’s deep tech and semiconductor ecosystem, what would it be?
(27:52) Madhu: Market linkage.
(27:53) Prakash: Beyond technical expertise, what’s one crucial soft skill for a processor architect or a chip founder?
(28:00) Madhu: I would say ideation.
(28:01) Prakash: One passion of yours outside of chips and building complex systems.
(28:06) Madhu: Oh, animal welfare.
(28:07) Prakash: Your ideal chai & chips combo while debugging a complex chip design.
(28:12) Madhu: I would say ice cream.
(28:13) Prakash: Okay. That works. So coming back to the first part of the conversation was more oriented towards what it will take, what are the gaps in terms of product conceptualization and bringing some of those expertise up in the market like India. The second part of this conversation I wanted to focus on InCore and what makes you guys unique, what’s been the unique value you’ve been able to add as it relates to the strategy that you were talking about. So the first part of this conversation would be InCore, you’ve got processor and IP focus. It’s based on RISC-V which is quite dramatic the progress you have made in the past several years. You build an ecosystem of creating reference designs for different segments and more recently you did System on Chip generator, SOC generator, which cuts cycle time from months to minutes. So, what makes InCore unique and how does it tie into the strategy that you have been advocating for strategic autonomy for India in semiconductors?
(29:26) Madhu: One word is probably not the answer you’re expecting. It’s empathy for our customer. Look, we started by doing cores and fabrics. Our USP at that point of time was we did generators so that you could do very customized cores and stuff. So it was a technical differentiator and quite unique in the world. So there’s no cookie cutter standard parts in terms of IP. You want more cache, you want an extra functional unit, all of that can be done, right? Add an instruction, all of that. So technically we are pretty cutting edge in that aspect.
(30:06) Madhu: Other things also, we used a very high-level language called Bluespec which apart from dramatically increasing productivity also dramatically increases quality because a whole class of errors cannot come when you do with Bluespec. For example, the Verilog it generates is guaranteed synthesizable. Correct? So there’s a whole host of tech. So InCore not only pioneered RISC-V cores, it also pioneered how you build a CPU, including building CPU talent in the country, right? So a lot of that. But that was a necessary but not a sufficient condition. If you look at how InCore has evolved, and we have evolved quite rapidly over the past just 18 months. We have gone from doing core to fabrics and security accelerators to creating SocGen, right? SocGen is just the tip of the iceberg because what InCore is doing essentially is solving the how do you build a SOC problem.
(31:04) Prakash: And what is SocGen, if I may ask?
(31:05) Madhu: SocGen is a tool that takes our RISC-V cores, our fabric, takes third party IP, and builds you an SOC at the digital level. Right? And it basically gives you FPGA output also so that you can immediately test the SOC on an FPGA. It generates some verification collateral which we’ll be adding more as we go along. It generates device driver stubs. Correct? We also have a process by which one time you can import any third-party IP. Correct? It also generates the full documentation.
(31:37) Prakash: And before you move forward, I would request you to share with our audience in simple terms, what is RISC-V? And what makes it different from every other CPU architecture that exists in this world?
(31:55) Madhu: RISC-V is just an ISA specification. Right? ISA basically is a combination of an alphabet and a grammar of how the computer at the lowest level computes data. If you got a high-level program in C or something, your compiler translates that into instructions and those instructions are the one that get executed. The instruction format is what RISC-V standardizes. Plainly put, if you look at it purely technically, RISC-V is no different from any of the other ISAs that have come across. Its key aspect is it’s open. Right? There’s no royalty, there’s no patents on that. You are free to do whatever you want. The x86 cannot be licensed because Intel and AMD don’t license it to other people. Arm is licensed, but it is proprietary, which comes with its own challenges. Technically Arm is good, right? Please don’t get the impression that I’m badmouthing Arm. It’s a great ISA. Arm has done a great job in creating this industry. But because it’s a proprietary ISA, that comes along with its own set of legal and business issues and so on. And if you want to for example, you’re doing a device, you want to change the ISA, you can’t do that unless you get an architecture license, that is expensive, right? So dealing with Arm can be a business issue. It’s not technically it’s a mature ISA, it’s a good ISA.
(33:10) Madhu: RISC-V opens it up and makes the ISA definition and use of ISA a lot more democratic. It comes with its own set of problems in the sense that potentially it can fragment and all of that and you have multiple choices and all of that. So early adopters need to have a strong stomach to use RISC-V. But now the industry has standardized. There are a lot of standard definitions, standard parts and all of that. So RISC-V, at least in the embedded segment is on par with Arm in terms of maturity, in terms of ease of use. So you can take a risk on, bet on RISC-V that is not more of a risk compared to taking a bet on Arm. There’s a pun there, but let’s ignore that, right? That is unintentional. When it comes to the higher end segment, RISC-V is getting there. I suspect in about two to three years RISC-V will catch up. And in some respects, it will probably exceed Arm faster because, very simply put, there are more brains working on extending the ISA compared to Arm. The ecosystem is much wider and deeper. Much wider, right? There are literally thousands of people working at problems. So RISC-V will throw up 20 possible solutions for a problem. That’s a double-edged sword. You have 20 solutions now to solve a problem. So I’m not downplaying the problem that causes, but if you’re an innovative company, then you really have choices, correct?
(34:39) Madhu: So for people who can leverage RISC-V, it’s a great thing. But that also, but if you come to somebody like us or other RISC-V companies, we will also narrow down the choices for you and give you something that you’re comfortable with. So RISC-V accommodates both type of companies. Companies who don’t want to make too many choices and stuff, “Okay, tell me what a safe approach is.” That can be done. Or companies that say, “Look, I want to experiment, I want the most cutting edge. I’m willing to put people into it and I know I’ll have problems and I’m okay with that.” It accommodates that also. So RISC-V is all about choice. Again, I’m not blaming Arm. A proprietary ISA necessarily has to restrict choice, otherwise you can’t. If Arm allows every customer to change the ISA, they would go out of business. It is the nature of being a proprietary ISA that is an issue and not because Arm itself is a bad company or something like that, right? So the battle is not between Arm and RISC-V, the battle is between a proprietary ISA and an open ISA. And the world will decide which works better for what circumstance.
(35:45) Prakash: So suppose you but you need not suppose, you actually have an open architecture. You have commercial ventures like InCore and several other chip companies coming up in India building on top of that. You have fixed or get better on product conceptualization and IP creation locally. What segments would you as InCore and other such companies focus on in India to create strong momentum and drive local consumption? You initially talked about let’s pick five or six designs and build volume. What would those segments be?
(36:26) Madhu: Actually we’ve written a paper on that and published that almost like two, three years ago. I got iSPIRT to publish it based on some work that I’d done. The segments are well-known. If you look at what we are consuming, we are consuming smart cards, we are consuming meter chips, we are consuming IP camera chips, we are consuming general purpose application processor chips that go in handheld terminals, Aadhaar terminals, panels and all that. This would be equivalent to an NXP. So you remember Intel had these StrongARM chips, right? Those category chips. So you’re basically talking 400 MHz to 1.2 GHz, quad-core with a little bit of display and video codecs and so on. BLDC controllers, standard control,mixed-signal chips, some automotive controllers. So there are like about five to 10 chips which drive most of the volume.
(37:18) Madhu: So, actually if you go to Arrow or Avnet or any of them, they’ll tell you what is shipping in volume, correct? It’s a very easy exercise to do. I did this exercise for the defence folks. I actually was an advisor to Bharat Electronics for about 18 months or so. So we actually sat down and did a detailed analysis. They were using 40 different processors across systems. Can you imagine? Ranging from an 8-bit to a 64-bit. Each one has got a different debugger, each one has got a different compiler and stuff and all that. It’s because you buy equipment from across the world, everybody has got their favourite chips, right? There’s nothing wrong. This is a problem that was thrust on BEL. BEL did not create the problem. BEL has got its own internal. So we found that we could, 80% of that you could map to just two chips that we created locally. And we started a program on that. Unfortunately, that did not see the light of the day. It’s called the D1 and the D2 processors. That’s there in my blog also. Okay. So, we found that we could match 40 CPUs just into two chips. We looked at all tanks, missiles, aircraft, everything, right? Most of the ones used in the strategic sector, not just defence, is all below 500 MHz. Okay. So that’s there, right? So those segments are very well-known. So, I often keep saying that India fabless should focus on chips costing below $10 first and hit that market.
(38:44) Prakash: So a follow-on in that is, suppose you identify those segments, whatever those segments are, right? How will the model work? What is the production or demand side of it? What is the supply side of it? Where does government play a role, if at all?
(39:02) Madhu: The government just has to do just one thing, which is demand creation. And we’ve been asking this for a long time. We need mandates for local procurement. The instant you put in a mandate, the VCs will fund it. The government does not have to fund at all. If you got sufficient mandates, you don’t need the DLI type programs for those categories. But if you want to do DLI, and MeitY has done an excellent job of doing the C2S program has got a BLDC program, right? That’s the right way to go. It’s not either-or. You keep doing the existing C2S, DLI and all that, that can continue, right? As the MeitY secretary once put it, he said, “Madhu, I want to make sure no startup complains 10 years down the line they’re starved for funds.” Of the 100 companies I’m funding, most of them I’m expecting to not make it. That is perfectly okay. But I want to make sure starvation of funds is not the reason they failed. That’s a very enlightened approach take, correct? So that will go on a tangent, correct? But if for, if we identify five chips and have a focused DLI only for those five chips, I would go so far as to say that why not get for example, C-DAC to do two of those chips based on our core, somebody else’s core and just license the GDSII to 10 people and say fabricate.
Prakash: And they could be anywhere in the world because fab may not be that much of a value-add versus before that in the entire value chain, right?
(40:33) Madhu: So there are multiple approaches. By no means am I preaching. These are just ideas coming from me. There may be terrible ideas. That’s for somebody else to judge. It’s some options that I’m offering, that’s about it. Take it for what it’s worth.
(40:49) Prakash: And in that model, suppose you consider five segments, you have DLI for that. Just to play this out completely, and some of your past conversations, you have been frustrated saying it is not for government to solve everything, it is for private sector to take the lead, and I completely agree with that. And at the same time, every other part of the world where semiconductor ecosystems have evolved, they have evolved because of active and I would say proactive state support. So what’s the balance that the government can play in terms of a role? What kind of intervention, nuanced intervention can government do to really keep a balance on driving local consumption, driving volume, but at the same time not really stifling private innovation and giving rise to sometimes trade disputes? So, what’s the balance you think we can have in that model?
(41:49) Madhu: We’ll have to learn that as we go along. That’s like saying what is the balance that works between two people in a marriage before you even met each other, correct? See, the key thing is no blame. Don’t blame anybody. Okay? Everybody has got the right intention. Don’t doubt people’s intention. Not the government, not the private sector, right? There is no enemy in this. Okay? Everybody wants to do the right thing. There are bound to be differences. We need to know how to resolve those conflicts, put them aside, and work for the better good. So that is where you start. Then you come on a common forum and say that look... but like I said, you need concrete goals. The goal I would like to see is a very simple goal. In 2028, can we ship 10 million microcontroller type parts out of this country? Designed in India, manufactured anywhere else, hopefully packaged in India, but need not be so. Pick a couple. See, I’ll tell you what the problem is, right? All of this is new. So there’s a great deal of fear. Once we get two to five successful chips, most of those fears will vanish, right? So all your fears would have been crystallized and you will know that some fears were not justified, some fears were justified. You will figure out how to tackle those fears which came true and the fears that didn’t come true, you realize that is just theoretical thinking on your part. So practice makes perfect. You and I can keep pontificating, but we need to get five separate chips out and then a lot of things will fall into place, the fear will go away.
(43:33) Prakash: Action is the only thing.
(43:34) Madhu: Action is the only remedy to get over this hurdle. It may fail, it may, all of that, right? Fear of failure, just ignore. Just go ahead with it and get it out. These are not complex chips. They will cost you what, 5-7 million dollars till mask creation. I’m not talking about placing wafer orders, right? Say 10 million dollars per chip, per chip.
(43:59) Prakash: So if you’re saying five or seven projects, so times...
(44:02) Madhu: 50 million dollars, 70 million dollars. That’s chump change for this country. So, if we spend 50-70 million dollars, we will have five good chips in five categories that can ship in volume in about three years. Why don’t we take that gamble?
(44:19) Prakash: Agree. And do you think the targets or the goals, one of them is definitely X amount of volume by this time frame. Great goal to have as a country. Are there non-numerical goals? The reason why I’m asking this is 5-7 years out, if you’ve achieved that volume, what else would take or make India have a stronger seat at the table in the semiconductor landscape around the world?
(44:49) Madhu: Actually the reason we are even at this stage is because the EMS companies, the unsung heroes of this revolution because see, they are the base, they are the foundation. Because now I can go to an EMS company and say that what are the components you’re procuring, can I do a substitution, right? See, again, just like we insult service companies, we also insult EMS companies. “You guys just do soldering.” It’s no value. No, no, but it’s a culture, right? Our maid doesn’t add value, the guy who cleans the garbage doesn’t add value and stuff. Society is built on these bases, on service companies and EMS companies, our maids, our garbage collectors, these are the guys on whom the society rests, right? I mean they should be praised to high heavens. We don’t.
(45:37) Madhu: But the EMS companies have built a great base for us where we have a manufacturing base, right? Because now I know what products are getting manufactured, I know what products are getting sold in the country. So I would pick two, three categories, LCD TV, mobile phones, inverters, the whole power category, and do a BOM analysis and figure out what are all the chips we need to localize. And then figure out what is the strategy to convince them. See, in some cases the government can mandate, but the government can’t mandate everything, right? So, we need to have a discussion with the EMS companies on localization of the parts. Which is already happening, that can be expedited.
(46:14) Prakash: Fantastic set of recommendations. I would agree. So, switching gears, you talked about or laid out a great set of next steps we could drive towards to build a very strong foundation for chip ecosystem in India. What I would also want to talk about is looking ahead. RISC-V as a platform is a great equalizer in times of tariffs, all the challenges that the whole world is facing, and the US-China tension. But you’ve also spoken about RISC-V chips including verification of those chips to a large part is done outside of India, specifically by Chinese companies. So how would you think about the absolute digital sovereignty, semiconductor included for India as it relates to RISC-V? Does RISC-V platform offer the mitigation to the challenges and how exactly should India approach it to get to the absolute digital sovereignty?
(47:22) Madhu: Well, this is part of the work I do for the Government of India. I advise the some of the strategic folks in the defence sector and so on. So we’ve done a fair bit of analysis on that. See, we have a fairly good idea as to, first you have to decide where your strategic threats lie. Correct? And we need to figure out how to build talent to build those chips. So my recommendation always is that a country our size should have local manufacturing at least at 30% or 40%. You should never have a sector where you are import-heavy. I’m not a firm believer that you need to be an autarky, that you close the borders and stuff. Trade is useful, trade needs to be there because if you completely shut off, there’ll be everything from price gouging and people taking it easy. So the local industry needs a right balance of threats from abroad, but with the sufficient amount of support. That that balance is very tough to achieve. And it’s a dynamic balance, right? You need a very proactive government.
(48:31) Madhu: So even in the, when I talked about when I was talking to MeitY for these five chips and all of that, I saying, “Look, for nearly 100, for nearly ever since independence, the business guys have always been telling the government, please stay out, right? You guys are intruding too much. But for once, I’m begging you, please intrude.” I like working with MeitY, right? They always say, we are scared of doing prescription to the industry and stuff, right? They’re very non-intrusive in the industry and I’m going the opposite saying, please be a little more intrusive, right? Because the industry is nascent and the industry doesn’t know what to do. This is one place where the government probably needs to handhold, at least for the next 10 years. I’m not saying it has to be a permanent thing. So like I said, there’s no blame. Everybody’s trying to figure out what is the right amount of government support, correct? And there’s no answer to it. Nobody knows the answer.
(49:27) Prakash: And every country is different because. Yeah, every country is different, history, World War II, all of that and that fuelling of semiconductor industry happened very differently and may demand a different set of interventions as compared to India.
(49:40) Madhu: Yeah, there’s no, there’s no standard prescription, right? So that’s why I prefer to take a practical approach and say that look, set practical goals, work backwards as to what will work to get those things across. But once the industry is producing 100 to 200 million chips, then the government can just back off and just give support and all of that, correct? But we need to identify, your question was about strategic autonomy. We need to figure out what do we mean by strategic autonomy. Correct? So does that mean chips for LCD televisions, which is a huge market? Chips for FMCG or not FMCG, white good appliances, mobile phones, correct? So we probably need to take segments where there’s a significant import component and then figure out how to localize those and see whether we can increase chip production because once you got volume, it’s very easy to justify local designs. You don’t want to get into... See, one problem we’re going to have is DRAM and flash, correct? So you can say that look, somebody can do an embargo on DRAM and flash, but my point is there’s nothing much you can do about it. Breaking into the DRAM and flash market is not for the faint of heart. I mean, I don’t know whether people know, flash and DRAM gets sold on the stock market. You can go to dramexchange.com, you’ll get spot prices. So even when I hang around in, you know, these Ritchie Street places, all the DRAM modules are kept inside the cash, next to the cash drawer with the owner. That alone is not kept outside.
(51:02) Prakash: I did not know that.
(51:03) Madhu: Yeah, only that is kept inside, next to the cash drawer. He keeps a very close watch on it and the prices quoted are valid only for four hours. If he quotes in the morning, it’s only till lunch. If you want to buy after lunch, you’ll have to get a different quote because the prices fluctuate during the course of the day. There are people who buy RAM and flash as a commodity and trade it. They are not techies, they’ll just buy so many terabytes of RAM and stuff, just buy and sell on the stock market because the PC guys when they suddenly have a demand and they can’t get it from the factory, they’ll buy from the spot market. So it’s a pure commodity market, which means you can take years of losses. OLED is in the same category, flash is in the same category. Flash I think is finally starting to make money. We had a bit of a price downturn. So you just need to look at NAND flash prices, right? You’ll realize that it’ll never follow the same downward price curve like a typical technology product, right? It’s like oil and pork futures or soybean futures.
(52:04) Prakash: So it does require a lot of orchestration at the highest level and then down to the next level detail and then all the way down to which segments, what volume, what time frame. So yeah.
(52:15) Madhu: It’s not a simple thing. I’ve been talking to somebody to see if we can get some white papers and stuff going on on this, but then InCore is a full-time job, so somewhere hopefully if I find some time, we can write a little more on this because I don’t think people realize how complex this whole industry is. Everything from the... it’s not Semicon, right? It’s a component industry and all of that.
(52:39) Prakash: Yeah. And frankly, despite being in the semiconductor industry for a couple of decades, I must say that since I started this podcast, I’ve learned so much about the industry that I did not know, frankly.
(52:56) Madhu: Yeah, if you’re sitting inside a particular company, I think you’re at Intel for a long time, right? You’ll have a very narrow perspective depending on your role, correct? You need to be outside. I’ve been fortunate that I’ve been kind of been a floater, running a whole bunch of startups. So I’ve worked with the EMS companies, I’ve worked with component guys, I’ve worked with distributors, right? So if I have a pricing question, I’ll just call up Arrow or Avnet or one of those guys saying, “Hey, what do you think is a fair price for the part?” and they’ll tell me immediately. I don’t have to do detailed market analysis. It’s two calls and you’ll get your validation.
(53:32) Prakash: So on all that you laid out makes immense sense. The question I have though is that you’re trying to move this at a time when there are big distractions. One of them is AI and second is as it relates to AI, the chips that go into AI. Then you get into leading edge fabs that drive some of the volumes on these kind of chips. So how do you compare and contrast that versus 5nm, 7nm foundries, AI chips versus the strategy that you’re talking about?
(54:07) Madhu: No, like I said right at the beginning of the talk, uh you need a strategy for each vertical. We do need a foundry strategy. I’m not the expert on foundry, so I will desist from making any recommendations. My only point there is focus on everything from 300 nanometers to 40. Maybe 28, 22 because that’s where the bulk of the demand in India is. So if you need autonomy in this country, you better have the older node fabs. The leading edge fabs depends on what kind of a strategy you want, how much you think you will not face an embargo from say from Taiwan or something.
(54:45) Madhu: But what are you going to build in those leading edge fabs is the other question that you got to ask. So what is our AI/ML chip strategy going to be? Are we going to focus mostly on training and stuff and basically say that look, sorry, for training we can get from NVIDIA and stuff, inference is a huge market so inference chips is what India needs to build and let’s not worry too much about training. That’s an approach, I’m not saying it’s right or wrong, right? Then we need to have at least two, three companies doing inference chips, both at the lowest level, the mid-level and at the highest level.
(55:17) Madhu: The second question is we are building all these data centers like crazy that are coming in India. How many of those do you want to have local chips? Can we have Xeon competitors to in RISC-V. There’s no reason you can’t. But then, then you need to take a look at what workloads are they doing. Correct? So I’m a firm believer in having simpler, larger core-count chips. So which was the SPARC approach. So I don’t like too heavy a processor but that requires that you write a lot of multi-threaded software. It places a much bigger burden on the software guys. But it gives you far more throughput, it’s more energy-efficient, right? So we need to have a chat with the data center guys to figure out what workloads they are facing, what kind of chips they want. Because all the other mundane chips also get sold in a data center, right? The networking chips, the regular old multi-core, 16-core, 32-core, 64-core CPUs which can be done with RISC-V, they’ll be competitive with. Look, you will not reach the maturity of a Xeon or an IBM POWER anytime soon. But high-end RAS and all of that, right? You start from the bottom. You look at non-critical workloads, you look at cloud workloads which may not be, not exactly banking workloads and all that, right? So I’m part of a working group for the government which has been looking into that.
(56:47) Madhu: But you need to segment the problem and say that look, I want to do this, but more importantly, you need to say I don’t want to do this. It is far more important to decide what not to do than to decide what to do. You need to determine first what you don’t plan to do. Because otherwise that’s a distraction. Then your search space becomes much smaller, then you can prioritize. If you have not eliminated what you don’t want to do, then it’s kind of tough to move forward.
(57:16) Prakash: And that’s really important because when you consider volume and I completely agree with you, the 180nm, 65nm, the mature nodes essentially have so much more volume than the leading edge nodes. So if you want to drive volume, if that is your priority and build local ecosystem, then I agree, your approach of five segments, those segments what you highlighted make absolute sense.
(57:42) Madhu: Look, we can learn from the aerospace guys, right? Nobody designs a plane and an engine simultaneously. So if you’re trying to build a semiconductor ecosystem, you don’t want to experiment with leading nodes because then when you’re building a semiconductor product, you’re experimenting with the product definition as it is. You don’t know whether the product is going to succeed or not. Why do you want to add the burden of playing with a leading node also simultaneously? So you need some invariables. You need constants.
(58:11) Madhu: The reason languages like Rust, I just want to jump into software, functional program works so well because you say X and Y are constants, they are immutable. So you don’t have to reason about them. They are guaranteed to stay. You worry only about these three variables which change. So all your verification will touch only those three variables. The problem with C is anything can change under you. So if your entire substrate under you is unstable, your foundation is unstable, you go haywire. So human beings can deal only with X number of variables simultaneously. That’s how we are constructed, correct? Too much of choice is bad for us. Mutual funds is a classic example where it is supposed to make the problem easier, but we have more mutual funds and stocks now, right? So it kind of defeats the purpose. It’s easier to pick a stock than a mutual fund sometimes. Correct?
(59:04) Madhu: So in the semiconductor strategy also, don’t make everything a variable. You’re not going to bring us on par with Taiwan in a year. Okay? We go at our own pace. Stop looking elsewhere. Don’t compare yourself with China, Taiwan, whatever, right? It’s like just wear blinkers, set a strategy, believe in yourself. Aatmanirbharta is not just doing stuff locally, it’s believing in yourself. People miss that aspect of it. It’s faith in yourself. Right? The faith has to come first. The local products will come next. Right? So that belief in ourselves. See, there is a fine line between strong belief in yourself and arrogance. Right? So we need to figure that out. So you believe in yourself, you take a gamble. Don’t be hard on yourself. Don’t critique ourselves so much. As a country, we have this habit of self-flagellation. We are a nascent country. Of course, we’ll get a few things wrong. Ho gaya, what has happened has happened, ignore. Stop blaming people. Right? What do you do next? So even in semiconductor, don’t try to invent everything under the sun, right? Let’s look at the old mature nodes. Let’s let’s figure out 10, 20 chips that we can do. Because once you’ve got an ecosystem going, then the confidence will come. And you can build from there. Then you can build from there. That doesn’t mean that you ignore the high-end nodes, right? You have a few R&D programs where you will build 2nm chips for high-end training and inference. Keep that going. So there has to be a combination. It can’t be either-or. But where you have to put X amount of money, where you need to put Y amount of money, you got to decide.
(1:00:39) Prakash: Agree. And all of these, at least based on my experience, are long games, Madhu. Like 10 plus years, couple of decades game, right? All the guys who are right on top, whether it’s Silicon Valley, Taiwan or China, they have done this for like two, three decades consistently. So, yeah, you alluded to it but I thought for the audience, it would be important to understand that this is not a short-term game.
(1:01:06) Madhu: It’s not a short-term thing, right? And especially the senior folks in the industry, that of course includes me, need to let go of our ego. So if you’re building a wall, you should realize that in your lifetime, you’re going to build three layers of bricks. That is your job. You build those three layers and you let the next set of folks build the next three layers and it goes on. Right? You don’t be arrogant and saying that I’ve got this 30-year vision. No, you can’t have a 30-year vision. That’s nonsense. You should have a good grasp of what you can build, what you’re trying to build. You build that foundation and, and the design will keep changing. See, I think I wrote in LinkedIn saying that CEOs especially have this God complex and society forces that. We need to let go of that. What we need is a large dose of humility. And then things will happen.
(1:02:05) Prakash: So well said. And that also brings me to another question which I wanted to ask you for the aspiring founders or the current founders in the chip and deep tech domain that are listening in as part of the audience. You have shared so much about the chip and deep tech industry, the difficulties and sometimes it can be really daunting to build such companies, including the ones that you are building, InCore. So what advice would you have for current founders or aspiring founders who are interested to building chips and deep tech ecosystem for India?
(1:02:42) Madhu: Know your market well. That’s all, right? Really study the market, understand it. There is a danger there. If you’re too optimistic, you will overread the market. If you’re too pessimistic and people discourage you...So I’ll tell you for example, right, classical example. Commodity chips. Chips selling for 50 cents to a dollar and a half. Should a startup build one or not? What do you think? Right? There’s no easy answer. The standard answer is you can’t differentiate, the big guys will, because see, commodity chips are a cash flow problem. There’s nothing tech about it because a lot of people can build that chip. Everything from inventory to market fluctuation, to canceled orders, to time of payment, to receivables, these are all the things, right? It’s not for the faint of heart. But that doesn’t mean I discourage people, correct? So if you’re going to be in the commodity market, you need an angle which differentiates you. If you can find that, get into the commodity segment. But if you can’t find that angle, don’t get into the commodity segment.
(1:03:48) Madhu: See, people still open rice mandis. You can still go put a small booth and sell Fanta. It’ll still sell. Right? Business is business. You can always do business. Correct? You need to figure out how much it costs to fund, cash flow, whether you can make a break and all of that, right? Look, I closed down a lot of startups that I tried because I was a naive idiot. There’s no other way to put it. Thinking that unless I am perfect and I am the best in the world, a startup won’t succeed. That’s like absolute nonsense. So everything I did when I initially did startups, I wish I’d not done because I did a social media startup, we did a medical records management company, all of those could be, could have taken and become big companies. But building startups is just hard work here. It’s 99% hard work, right? The differentiation and all that fancy stuff is like 2 to 3% max, correct? It’s a lot and lot of hard work. That can be done. But you need the patience. But which also means that you need investors who will be patient with you. Right? That probably is a lacking commodity. See, a lot of businesses are good businesses worth investing in, but they may not necessarily be a VC investable. People unnecessarily, that’s one more group that is vilified unnecessarily. Right? Seems to be the trend for today. It’s like VCs have a particular business model. They expect a certain return. So if you fit within that model, they will fund you. So if you fit within that model, you’re VC investable. If you don’t fit in, doesn’t mean you’re a bad company. You simply don’t fit in with what they’re expecting. So if a VC rejects you, you part in good terms saying that look, we are on different trajectories in life. That’s all. End of story. That’s the same, right? It’s like we waste so much energy in blaming people. That’s like, it’s not funny.
(1:05:50) Prakash: So we are coming to the end of the show, Madhu. So I have a final question which I ask all my guests. The India’s journey on semiconductors and deep tech is not only about domestic growth, it is putting India on the global map on technology. So if you fast forward a couple of decades or so, what is your vision of India’s position in the technology world? And what needs to happen today in order for us to achieve that?
(1:06:23) Madhu: See, I’m a hard numbers guy, right? So for me, uh India probably needs to have 25% of the world’s market in semiconductors costing below $25. For me it’s a 25 by 25. One fourth of the market should be chips designed in India. Fabricated, all of that, I can’t figure out, but quarter of the world’s chips should come out of India. Below $25. Above that, kind of tough depends on how everything and goes, but in the higher end chips like servers and all of that, mobile phones and all that, I would like something like about 15% of the world’s market. In terms of value, I would aspire to about 20 to 30% of the revenue in the semiconductor world should come from India. Okay. I think that’s a fair... in keeping with our... Hamare aukaat ke bahar nahi hai, to put it in Hindi.
(1:07:21) Prakash: And what needs to happen? You talked a lot about what the strategy should be. Anything you missed in terms of what needs to happen today to achieve those goals?
(1:07:29) Madhu: Each of those segments like we said, right? We need to figure out local champions and then measure. So if I take an LCD television apart, I should say, look, next two years 20% localization. So the government has thought through all of this, right? PLI is also tied to PMP. But the PMP is lagging. So we need to crack the whip on PMP. Mobile phones, we are, we are what, we are shipping about 120, 150 million mobile phones a year. 90-95% of the components are imported. Look, take a bill of material on a mobile phone here. Ignore DRAM and flash. Other than that, why can’t all the other components be local? What prevents you from it? At least in the sub 10,000 rupee segment, we should be able to do. You will say yeah, modems take time to develop. Okay, license a modem. Fine. You can license modem tech, should be available. Other than that, everything else is doable. What is the part that you can’t do? You can’t do audio codecs?
(1:08:40) Madhu: So interestingly, you’ve seen Axero, right? You’ve heard of Axero?
(1:08:44) Prakash: No.
(1:08:45) Madhu: So the Murugappa Group, which has put the OSAT, went and bought the entire RF division of Renesas. Lock, stock and barrel including the CEO. Very interesting chap. I’ve talked to him. You should look at them. They’re nice. So now India is one of the world’s leading supplier of RF parts. To Nokia, to Ericsson and to Huawei. So Chinese telecom now runs on Indian chips. So I think they, they asked them to put an R&D unit in China also. I saw Axero had opened a design office or something in China. I’m pretty sure the Chinese are getting nervous and will try to swap us out and stuff. But as of today, India is one of the leading suppliers of RF chips to the world.
(1:09:31) Madhu: Yeah. So one of the things that Indian companies need to do is a lot of distress in the semicon industry around the world, right? We should be grabbing companies left, right and center. I have sent two, three proposals to the government also. There are a lot of fire sales going on. This is the right time to pick up IP.
(1:09:46) Prakash: In fact, interesting you say that because just in today morning’s Mint, there’s an article about how Indian companies could seek out as a part of the chip ecosystem development exactly what you’re saying. IP around the world and why don’t we invest in them?
(1:10:06) Madhu: Look here, it’s like, see, ultimately you’re either a semicon guy or not. If I look at a nice chip, preferably an InCore IP chip, but third party is okay, when you, when you look at a printed circuit board and you run your hands on the chip, there is a deep sense of satisfaction. Look, at what we have created in this industry. We pack in trillions of electrons and create a virtual universe, all from grains of sand. If that miracle doesn’t give you spiritual satisfaction of some kind, you are dead inside, man.
(1:10:44) Prakash: Yeah. 100% agree. 100% agree. As someone who has spent pretty much my entire life in silicon, there is no better industry to be. It’s one of the most fascinating and magical things invented by humans.
(1:11:00) Madhu: It’s as close to magic as you can get and I’m thrilled I was given the opportunity and absolutely no regrets. It’s been a wonderful journey. So there’s been no downsides. Little things, yeah, revenue going up, down, difficult customers. Those are all trivial things, right? You need to, at least at a certain level of maturity, you need to figure out what should bother you and what should not. Otherwise, you let the little things bother you, you’ll never make progress.
(1:11:34) Madhu: The only thing I would say is the veterans in the industry need to speak up more just to provide support. That includes me. I’m including myself also. I think the veterans need to support the industry more, talk a little more, do something more concrete because there is still a bit of a crisis of confidence. I still see startups sinking. Startups with good products. There’s no business to sink. We kind of need to go out of our way to help them. I think that can be done. Look, it’s, it’s always think big and implement small, correct? I think we’ve done enough of the big thinking. At some point of time, well, I have a bit of an OCD also. So for me, the details matter, right? You sweat out the details, everything else takes care here. So it’s like big, inspiring words and all of that is fine. Sweat the details out and the rest will take care of itself.
(1:22:33) Prakash: Very well said and that’s one of the reasons why in a small way I want to contribute to the semiconductor and deep tech industry here, Madhu. The reason why I started this podcast is to bring a voice into the industry, right? What, what people think and what founders like you have built and are continuing to build. It’s such an amazing place and more people need to hear the stories and ideas about what we can do to build the India’s ecosystem.
(1:22:59) Prakash: Thank you Madhu and for our audience, please tune in to whichever channels you listen or watch our Chai & Chips episodes. Please do give us feedback on how I can do better, what kind of topics you would like to listen to and what areas are really piquing your interest. Thank you once again.